home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!ogicse!flop.ENGR.ORST.EDU!gaia.ucs.orst.edu!umn.edu!mmm.serc.3m.com!us275168
- From: us275168@mmm.serc.3m.com (Kenneth A. Kiefer)
- Newsgroups: comp.databases.ingres
- Subject: Re: DEC ULTRIX INGRES checkpoint problem
- Message-ID: <1992Nov19.031726.12353@mmm.serc.3m.com>
- Date: 19 Nov 92 03:17:26 GMT
- Article-I.D.: mmm.1992Nov19.031726.12353
- References: <1992Nov13.175814.27160@aplcen.apl.jhu.edu>
- Organization: 3M Company
- Lines: 63
-
-
- mae@aplpy.jhuapl.edu (Mary Anne Espenshade) writes:
-
- > .... I've probably made 30 checkpoints with no
- > problem (and have even recovered from the checkpoint a few times). Last
- > week I ran a update and made the checkpoint as usual, neither process
- > reported any errors. When I next attempted to access the database, the
- > sql command aborted with the following error message:
- > E_US0010 Database does not exist.
- >
- > Going to the error log explained more of
- > the problem. For each attempt to access the database, the following
- > series of errors was logged:
- >
- > E_DM923C_CONFIG_FORMAT_ERROR Error in configuration file format.
- > E_DM923A_CONFIG_OPEN_ERROR Error occurred opening configuration
- > file, (aaaaaaaa.cnf).
- > E_DM9265_ADD_DB Error occurred while trying to add a database.
- >
- > I didn't want to try to recover from the checkpoint. I didn't know
- > whether it would be able to find the database or if the checkpoint
- > itself was corrupted. I finally decided to try replacing the
- > aaaaaaaa.cnf file with a copy from the previous day's system backup.
- > Since this is a binary file, I had no way to determine what had changed
- > between the two. This seems to have worked, since I could then access
- > the database and successfully ran the scripts that read the data.
- >
- > Hoping that the error was a one-time fluke, I ran the next tape update
- > and made another checkpoint, again with no errors reported. The exact
- > same thing happened. Once again I recovered from a copy of the
- > configuration file I had made just before I took the checkpoint (being
- > paranoid about such things), but I am left with no way to checkpoint
- > this database.
- >
- > Two alternatives have been suggested by the system maintainers - copy
- > out all the data, recreate and reload the database and start over
- > (assuming that the error had something to do with the number of times
- > checkpoint had been run), or stop doing checkpoints and just tar the
- > database data files to tape (since the checkpoint is only really
- > important if journals are tied to it).
- >
- > Has anyone else run into this problem? Is it safe to practically
- > ignore it this way or should I be doing something else.
- > --
- > Mary Anne Espenshade
- > ARPA: mae@aplpy.jhuapl.edu
- > UUCP: ...!allegra!mimsy!aplcen!aplpy!mae
-
- Here's a hunch. If you're checkpointing to tape, try including
- the "-d" flag on your ckpdb command to erase the checkpoint
- timestamp history kept in the .cnf file. Beware, if you're
- checkpointing to disk, the -d will also delete your old checkpoints from
- disk, so use it judiciously.
-
- The .cnf file stores various pointers and other such stuff about the
- database, including the dates/times of the most recent 16 checkpoints.
- As I remember, Ingres 6.2 has a bug where if more than 16 consecutive
- checkpoints are done without the "-d" flag, the checkpoint history
- overflows, clobbering other data in the .cnf file.
-
- Good luck,
- Ken Kiefer
- kakiefer@mmm.com
-